Human pulse recognition based on wavelet transform and BP network

Pulse diagnosis is one of the most important methods in traditional Chinese medicine diagnosis. In order to classify the human pulses effectively, feature extraction and pattern recognition are needed. In this paper wavelet transform is used to the feature extraction and the third layer wavelet energy and the wavelet entropy values are selected as the wavelet domain features. Therefore a new feature selection scheme is proposed, which combines wavelet domain features, time domain features and frequency domain features. Firstly, three kinds of features are extracted from a human pulse signal, and then the total 13 extracted features are fed to the BP neural network for pattern recognition. We name this method WBPN. The proposed WBPN method is tested on 12 different kinds of human pulse signals, and the average recognition rate of untrained test samples reaches 98.45%, which is higher than most of the existing methods. The results demonstrate the effectiveness of the proposed WBPN approach in human pulse recognition.